EPL 2013/14 Mid-Season Analytics

Just as the final half of the most fiercely competitive season of Premier League history gets under way, we at ‘Goalden Times’ built our own prediction model and tried to envisage the possible champion of 2013/14 season.

Predictive models use historic data to identify and quantify patterns and trends that can be used to predict future behaviour. Let us see how we can predict the champion of this season’s English Premier League based on half-yearly performance trend of last 10 years. Last decade was clearly dominated by the big houses – Manchester United & Chelsea FC with some rare success for Arsenal & Manchester City. But what makes these teams successful? We did a year-by-year analysis covering as many as 20 different parameters that could indicate performance and likelihood to win the competition.  Before going into details, let’s have a gaze over current league standings.

.

pyramid1

.

Analysis of English Premier League – Methodology

To implement the predictive analysis following steps are followed –

  • Analysis of match-by-match historic data for past ten years (since 2003/04 Season) and defining numerous key parameters that could indicate a causal relation towards winning the league.

.

  • The following parameters were defined and their contributions towards identifying league leaders were measured based on historical statistics. The results are as follows – [1]

param1

  • All these parameters were then assigned a strength based on their ability to identify the league winner. The worst prediction by any of the parameters was a team finishing fourth in the league table, in any season under our analysis.  If a parameter predicted winner correctly, then it was assigned a weightage of 0.7. For being runner up, weightage comes down to 0.3. Weightage even becomes negative, -0.3 and -0.7, for wrongly predicting a team as champion which happened to finish third or fourth in the final standing. By this practice, five strongest parameters were identified which will be the key metrics driving the predictive model.

.

  • Some of the parameters have shown some interesting characteristics – like the last one called “Rough Play”. It reiterates the fact that the winning team should have a killer instinct, they should be willing to commit tactical, or time-to-time genuine fouls (taking one for the team). However, out of top four teams, often the third and fourth placed teams committed less fouls, or were booked or sent off lesser times. Making “Rough Play” a weak indicator for champions!

param2

  • Then the values of these parameters were calculated for all clubs during mid-year (December end) and tested them on mid-year data to see how accurate these parameters were in the past. We have found an accuracy of 80%.

.

table

 …………………………………….

testing

    • After analyzing each team’s performances until 16th gameweek of 2013-14 season, the same parameters predict Manchester City scoring maximum and most likely to be the winner of 2013-14 Premiere League and the forces are heavily in favour of them. Liverpool and Arsenal are second and third favourites to win this trophy. In spite of the fact that Arsenal are actually leading the table right now, they are much less likely to win the trophy.

 first_half(1)

result

Final Prediction: The aforesaid Prediction Model and Analytics of last 10 years data indicates that the probable winner of this season will be Manchester City.

Football is a special game and we know that one moment of magic, one piece of individual brilliance or one wrong decision can turn a result upside down. We can only validate the accuracy of the model at the end of this season. Till then, enjoy the beautiful game.

.

Note [1]: The parameters that were used in our predictive model are explained below
  1.  Home Goal Ratio: Goals scored / goals conceded per home match
  2. HT Retention: When a team was able to hold its half-time result for the second half while winning.
  3. Win ratio against other top clubs: How many matches a club won out of total played against other top four clubs.
  4. Away goals conceded / match: Goals conceded/away match.
  5. HT Performance: Goals scored before half-time/ match for both home & away
  6. Away Goal Ratio: Goals scored / goals conceded per away match
  7. FTHG/match: Full time Home goals scored per match
  8. FTAG/match: Full time away goals scored per match
  9. HT Reversal: When a half time result was reverted resulting in either a win or a loss
  10. Home Goals Conceded / match: Self explanatory
  11. Propensity to convert to win - All those matches where the club was able to convert a halftime lead / stalemate into a win by fulltime.
  12. Team Accuracy - How many shots are on target for a team out of all shots taken
  13. Rough Play - A negative attribute which counts how many fouls, Red & Yellow cards were conceded by the players of a team.